Crowd wisdom and bias
Averaging removes noise. Does it remove bias?
Playback is compressed; model values are shown separately.
Make a prediction
As this crowd grows, what can averaging remove?
Ready. Adjust the settings or press Run.
How it works & assumptions
Each estimate is 100 + bias + Normal(0,20²). A crowd mean has error RMSE √(bias²+400/n). Compare individual estimates with crowd means.
Independent normal errors with the same fixed additive bias. Shown trials are random samples; finite-run errors vary.
Random runs are reproducible from their setup links. Replay keeps the same seed; New random run changes it. A sample run need not equal the theoretical average.